Hi Lucas, On 22.03.2012 [02:28:17 -0300], Lucas Meneghel Rodrigues wrote: > On Thu, Mar 8, 2012 at 8:36 PM, Nishanth Aravamudan > <n...@linux.vnet.ibm.com> wrote: > > Hi, > > > > So one gap in the current implementation of the install server > > functionality I noticed is that the end-user still has to fiddle with > > the backend cobbler server when they want to install a different than > > the currently selected profile. It seems like the end user should be > > able to select the profile to use on a host-by-host basis. > > > > Some implementation notes: > > > > - The link between autotest and cobbler is established by the label of > > the autotest host being found in the comment field for the profile in > > cobbler. > > I gave a look at the patchset, and I do like the fact that you > minimized intrusion in the current code. I need to go over it again. > What I have done is, pick up all the patches you've sent (including > the couple of corrections you've sent later), and made a pull request, > so I could review the patches and apply static code analysis in it > more easily:
That makes sense to me. > https://github.com/autotest/autotest/pull/240 > > So you're welcome to open your own pull request, that I could then use > as a base for reviewing and do all the back and forth necessary to get > this feature upstream. I consider it a good proof of concept, and I'd > like to work with you to get this upstream. Thank you! Yep, I plan on sending a regular pull request in the future, maybe later today. > > - Many spots still need to be parameterized on the presence of the > > xmlrpc URL in the global config > > I've noticed we could factor that code out to a library function. > Also, we could have a similar function to verify whether the install > server is up and running. Yes, that's a good idea. What is the best way to determine if the install server is up & running? I mean, we can use http requests to see if it responds, but that doesn't mean it's correctly configured. I guess, if we do a dummy remote get_systems or something, that may be sufficient? > > - In my environment, I always want to reinstall the machine before > > running any tests, so I've modified the default client and server > > control files to run machine_install. Perhaps this makes sense to add > > via a global config option -- always_reinstall_host or something. If > > that is added, then I think the verify-before-job job should be > > avoided if that option is set. Such an option also only makes sense if > > xmlrpc_url is set. > > Like we have talked over irc, I'm not sure whether such an option is a > good idea, but I'm not ruling it out completely. It might have value > to just install the test machines in a compulsory way, although it > seems excessive to me. In considering all the use cases autotest already supports ... I think it's probably better not to force this. So I'm going to add some changes to add a "<do not install>" option to the drop down box at job creation time. Thanks, Nish -- Nishanth Aravamudan <n...@us.ibm.com> IBM Linux Technology Center _______________________________________________ Autotest mailing list Autotest@test.kernel.org http://test.kernel.org/cgi-bin/mailman/listinfo/autotest
